Endnote: Estimating Income Inequality: Methodology and Assumptions
A. Methodology
There are two common approaches for estimating global income distribution (see
UNDP 1992 and 1995, Sutcliffe 2004 and Milanovic 2005). The first approach is
known as the inter-country distribution accounting model, which looks at the
average income differences between large groupings of countries. To do so, it
treats all members of a country as if they have the same income (e.g. all Bolivians
are assumed to earn the same amount of money in a given year). After ordering
all countries in the world according to their levels of per capita income (smallest
to largest), global income distribution estimates are derived by dividing the world
population into five equal parts (or quintiles) and calculating the corresponding
shares of total global income.
The data requirements for inter-country model are very basic and consist of GDP
per capita and population for each country. As a result, this method allows for a
very large sample size (about 98% of the global population for any given year)
and covers very recent time periods. All calculations are based on data from
World Bank (2011).
A second approach accounts for both inter- and intra-country distribution.
Frequently referred to as the global distribution accounting model, this method
decomposes national income by quintiles and compares those incomes across
countries. Here, the average per capita income of those in India’s bottom quintile
is estimated on the basis of their share of total national income. While this
method still assumes that large numbers of individuals have the same income (e.g.
a quintile of India’s population equals the entire population of Indonesia), it
allows for the construction of a hypothetical world in which all persons can be
lined up in a single distribution—within country population quintiles—regardless
of where they live.
The global distribution model has much more stringent data requirements than
the inter-country model. In particular, this method requires national income
distribution estimates, which are commonly presented as the share of total
income held by different population quintiles, from the poorest 20% (quintile 1
or Q1) to the richest 20% (quintile 5 or Q5). Annual quintile data were extracted
from World Bank (2011) for all available countries and years and then
supplemented by information from UNU-WIDER (2008) and Eurostat (2011).
Since we are most interested in understanding trends over unique time periods
(e.g. 1990, 2000 and most recent available), interpolation and nearest neighbor
imputation were used as gap-filling procedures to maximize the number of
observations using all three distribution data sources. We did not, however,
estimate quintile values for all countries in the world, which means that all of our
data points are derived from actual estimates.